svn commit: r179669 - head/sbin/rcorder

[Mike Makonnen]

Pawel Jakub Dawidek wrote:
> On Wed, Jun 11, 2008 at 02:24:14PM +0300, Mike Makonnen wrote:
>> Brooks Davis wrote:
>>> On Tue, Jun 10, 2008 at 10:01:13AM +0200, Dag-Erling Sm??rgrav wrote:
>>>> Mike Makonnen <mtm at wubethiopia.com> writes:
>>>>> Dag-Erling Sm??rgrav <des at des.no> writes:
>>>>>> How about s/REQUIRE/AFTER/, to mirror BEFORE?
>>>>> I would be in favor of that. Any objections?
>>>> (obviously, we'd have to keep REQUIRE for backward compat, and perhaps
>>>> emit a warning when we come across a script that uses it)
>>> I think we'd want to support REQUIRE quietly until 6.x and 7.x releases 
>>> that
>>> support AFTER have been shipped for some time and a sweep of ports has been
>>> done.  After that we could consider a warning in HEAD.
>>>
>>> That said, I'm not yet totally convinced (still thinking about it) given
>>> the level of churn involved.
>> Actually, I was thinking that we leave REQUIRE in there indefinitely for 
>>  backwards compatibility and avoiding gratuitous differences with 
>> NetBSD and DragonFly. It doesn't cost us anything but is a net benefit 
>> to our users. We can simply discourage its use in the documentation. As 
>> far as churn is concerned: I don't think it's that big a deal 
>> (especially if you use mergemaster -U). The patch is actually pretty 
>> simple (see attachment).
> 
> Just a note from a user perspective. Every time I install FreeBSD and I
> upgrade it to the latest release, I've to spend a lot of time to
> manually install all the changes like the above (I'm not used to -U
> option, but I also think it doesn't work for the first time.).
> And believe me, this is sooo huge PITA, especially if you have to
> install more than one system (but not many enough to automate it).

Which is exactly why mergemaster(8) has a -U option (it's such a 
time-saver I don't understand why it's not enabled by default).

> If there really have to be changes like that, please consider installing
> rc.d/ script as a part of the installworld target. More or less what
> kib@ suggested the other day.

The thing I don't like about it is that it would blow away any local 
modifications to those files. A lot of people are used to editing these 
files for major/minor local tweaks (in fact, I do it myself some times) 
and I think simply blowing these files away would inconvenience a lot of 
users.

> One of the things I was thinking about how to detect if the script was
> manually edited by the user is to keep md5 of the script in it. Of
> course we have to exclude one line. eg.
> 

Again:
why re-invent the wheel? The mergemaster(8) command already does this, 
except that the md5 sum doesn't live in the file, but in 
/var/db/mergemaster.mtree.

Cheers.
-- 
Mike Makonnen       | GPG-KEY: http://people.freebsd.org/~mtm/mtm.asc
mtm @ FreeBSD.Org   | AC7B 5672 2D11 F4D0 EBF8  5279 5359 2B82 7CD4 1F55
FreeBSD             | http://www.freebsd.org